AIPolicyTracker
Menu

AI incident ·

South Africa Draft National AI Policy Reportedly Included Fictitious References Believed to Be AI Hallucinations

4 news reports Snapshot 7 Sep 2026

In brief

An AI system built by Large Language Model Developers and Generative Ai Developers and deployed by Government Of South Africa and Department Of Communications And Digital Technologies (South Africa) allegedly harmed General Public Of South Africa, General Public and 2 others.

Risk domain
Misinformation False or misleading information
Occurred
Coverage
4 reportsApr 2026 - May 2026

What happened

South Africa's Draft National AI Policy, gazetted for public comment, reportedly contained at least six fictitious academic references. Several cited articles or journals reportedly did not exist or were disclaimed by journal editors, and experts said the errors were consistent with AI hallucinations. The Department of Communications and Digital Technologies said it was reviewing the discrepancies but argued they did not affect the draft's substance.

Laws that address this harm

Policy angle: Classified under Misinformation (False or misleading information) in the MIT AI Risk Repository taxonomy; 5 recorded instruments address this use case.

Matched from the record's risk domain and country to the instruments recorded here. A reviewer can correct the match in the repository (data/external/incident_overrides.yaml).

News reports (3)

Titles link to the original publisher; report text is not reproduced here.

  1. Govt’s draft AI policy cites fictitious references experts believe are AI hallucinations
    news24.com · William Brederode
  2. South Africa withdraws AI policy due to fake AI-generated sources
    reuters.com · Reuters, Nellie Peyton
  3. Fake AI references in South African AI Policy: Two men named in report
    mybroadband.co.za · Hanno Labuschagne

Who was involved

Alleged harmed party
General Public Of South Africa, General Public, Epistemic Integrity, Ai Policy Stakeholders

Classification (MIT AI Risk Repository taxonomy)

Risk domain
Misinformation
Causal entity
AI
Intent
Unintentional
Timing
Pre-deployment
Harm level
—
Sectors
—
Countries
—

Risk entries describing this failure mode

Entries from the MIT AI Risk Repository coded to subdomain 3.1.

  • Pursuing Consistent Context

    "LLMs have been demonstrated to pursue consistent context [129]–[132], which may lead to erroneous generation when the prefixes contain false information. Typical examples include sycophancy [129], [130], false demonstra...

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Defective Decoding Process

    In general, LLMs employ the Transformer architecture [32] and generate content in an autoregressive manner, where the prediction of the next token is conditioned on the previously generated token sequence. Such a scheme...

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Noisy Training Data

    "Another important source of hallucinations is the noise in training data, which introduces errors in the knowledge stored in model parameters [111]–[113]. Generally, the training data inherently harbors misinformation....

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Hallucinations

    "LLMs generate nonsensical, untruthful, and factual incorrect content"

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Faithfulness Errors

    "The LLM-generated content could contain inaccurate information" which is is not true to the source material or input used

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Factuality Errors

    "The LLM-generated content could contain inaccurate information" which is factually incorrect

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Untruthful Content

    "The LLM-generated content could contain inaccurate information"

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Knowledge Gaps

    "Since the training corpora of LLMs can not contain all possible world knowledge [114]–[119], and it is challenging for LLMs to grasp the long-tail knowledge within their training data [120], [121], LLMs inherently posse...

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

Incidents in the same risk subdomain

All incidents in this subdomain

Source record: incident #1467 on the AI Incident Database · all 4 reports